Upload LoRA adapter (README written by author)

README.md

@@ -35,8 +35,8 @@ while intermediate reasoning (Chain-of-Thought) is masked.
 - Method: QLoRA (4-bit)
 - Max sequence length: 2048
 - Epochs: 1
-- Learning rate: 1e-06
-- LoRA: r=32, alpha=64
+- Learning rate: 1e-05
+- LoRA: r=64, alpha=64
 
 ## Usage
 

adapter_config.json

@@ -29,17 +29,17 @@
   "peft_type": "LORA",
   "peft_version": "0.18.1",
   "qalora_group_size": 16,
-  "r": 32,
+  "r": 64,
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "gate_proj",
-    "q_proj",
     "up_proj",
-    "down_proj",
-    "o_proj",
     "k_proj",
-    "v_proj"
+    "down_proj",
+    "gate_proj",
+    "q_proj",
+    "v_proj",
+    "o_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d83b3b4617579cc0ca5729a929fac917429b8c5b0df5e191bf2aae32a786195b
-size 264308896
+oid sha256:8a9a90d0e8c98a219b162cfd15eb60eaea3264c77276ae961c212cc057362c03
+size 528550256